Rock shear strength parameter machine learning prediction method and system

Abstract

This invention relates to the field of rock mechanical property testing technology, and in particular to a machine learning method and system for predicting rock shear strength parameters. The method includes the following steps: determining rock shear strength parameter prediction indices and rock shear strength parameters, and establishing a sample database; improving the sparrow search algorithm using an improved Tent chaotic mapping, dynamic adaptive weights, Levy flight strategy, and Cauchy-Gaussian hybrid mutation mechanism; training a Stacking ensemble model using the sample database, and simultaneously optimizing the hyperparameters in the Stacking ensemble model using the improved sparrow search algorithm to obtain a rock shear strength prediction model; using the rock shear strength prediction model to predict rock shear strength parameters, and simultaneously using the SHAP interpretability method to evaluate the contribution of each rock shear strength parameter prediction indices. This invention can achieve high-precision prediction of rock shear strength parameters and effectively avoids the problems of long cycles and high costs in existing rock shear strength testing.

Description

Technical Field

[0001] This invention relates to the field of rock mechanical property testing technology, and in particular to a machine learning method and system for predicting rock shear strength parameters. Background Technology

[0002] Among numerous rock mass engineering disasters, shear failure is one of the main failure modes of rock instability. Cohesion and internal friction angle, as important parameters characterizing the shear strength of rock, have been widely used in rock mass engineering stability assessments. However, accurately obtaining cohesion and internal friction angle typically requires shear tests or triaxial compression tests. These tests consume significant manpower and resources for sampling and sample preparation, and are time-consuming and costly. This often results in engineers being unable to obtain relevant parameters in a timely manner during rock mass engineering stability assessments, thus affecting construction progress and safety. Therefore, how to quickly, accurately, and economically obtain these two rock shear strength parameters has become an urgent problem to be solved.

[0003] In recent years, ensemble learning algorithms, represented by Stacking, have demonstrated excellent performance in handling complex problems, but problems such as random hyperparameter selection still exist. Optimizing the hyperparameters of each learner model in an ensemble learning algorithm using metaheuristic optimization algorithms, such as the Sparrow Search algorithm, Whale Optimization algorithm, and Harris Eagle Optimization algorithm, can effectively avoid the instability of prediction results caused by random hyperparameter selection. However, these algorithms suffer from insufficient diversity during population initialization, unstable search efficiency, and a tendency to get trapped in local optima. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a machine learning method and system for predicting rock shear strength parameters.

[0005] To achieve the above objectives, in a first aspect, this invention provides a machine learning method for predicting rock shear strength parameters. The method includes the following steps: determining rock shear strength parameter prediction indices and rock shear strength parameters, and establishing a sample database; improving the sparrow search algorithm using an improved Tent chaotic mapping, dynamic adaptive weights, Levy flight strategy, and Cauchy-Gaussian hybrid mutation mechanism; training a Stacking ensemble model using the sample database, and simultaneously optimizing the hyperparameters in the Stacking ensemble model using the improved sparrow search algorithm to obtain a rock shear strength prediction model; using the rock shear strength prediction model to predict the rock shear strength parameters, and simultaneously using the SHAP interpretability method to evaluate the contribution of each rock shear strength parameter prediction indices to the prediction results. This invention can achieve high-precision prediction of rock shear strength parameters and effectively avoids the problems of long cycles and high costs in existing rock shear strength measurement experiments.

[0006] Optionally, the steps of determining the rock shear strength parameter prediction index and the rock shear strength parameter, and establishing a sample database, include the following:

[0007] Determine the prediction index and shear strength parameters of rock;

[0008] Collect data on predicted rock shear strength parameters and rock shear strength parameters;

[0009] A sample database was established using the collected rock shear strength parameter prediction index data and the rock shear strength parameter data.

[0010] Optionally, the rock shear strength parameter prediction index includes P-wave velocity, rock density, uniaxial compressive strength and Brazilian tensile strength, and the rock shear strength parameter includes cohesion and internal friction angle.

[0011] Optionally, the improved sparrow search algorithm using the improved Tent chaotic mapping, dynamic adaptive weights, Levy flight strategy, and Cauchy-Gaussian hybrid mutation mechanism includes the following steps:

[0012] Initialize the sparrow population using an improved Tent chaotic map;

[0013] Introduce dynamic adaptive weights into the sparrow search algorithm to improve the finder position update formula;

[0014] Follower position update formula based on Levy flight strategy to improve sparrow search algorithm;

[0015] During the iterative operation of the sparrow search algorithm, the Cauchy-Gaussian hybrid mutation mechanism is used to perform adaptive mutation operations on the top S optimal individuals with the best fitness values ​​in the current iteration step population, and the optimal individual is selected from the individuals before and after mutation to enter the next iteration step.

[0016] Optionally, the improved Tent chaotic mapping satisfies the following relationship:

[0017]

[0018] in, For the (i+1)th state value, Let N be the i-th state value, and N be the number of particles. It is a random number between [0,1].

[0019] Alternatively, the improved sparrow search algorithm's discoverer position update formula is shown below:

[0020]

[0021] in, Let this be the position of the i-th sparrow in the j-th dimension during the (t+1)-th iteration. Let be the position of the i-th sparrow in the j-th dimension during the t-th iteration. For the dynamic adaptive weights, Let be the global optimal solution for the sparrow population in the j-th dimension at the t-th iteration, and let Q be a random number following a normal distribution. This is a warning value. This is a safe value.

[0022] Alternatively, the improved sparrow search algorithm's follower position update formula is shown below:

[0023]

[0024] in, Let be the position of the i-th sparrow in the j-th dimension at the (t+1)-th iteration, and Q be a random number following a normal distribution. Let be the worst-case position of the sparrow population at the t-th iteration. Let be the position of the i-th sparrow in the j-th dimension during the t-th iteration. For population size, This represents the optimal position of the sparrow population in the (t+1)th iteration. The step size is random.

[0025] Optionally, the Cauchy-Gaussian mixture mutation mechanism satisfies the following relationship:

[0026]

[0027] in, This is the position after the mutation. For the first The position of the optimal individual in the j-th dimension at the t-th iteration. For random variables that satisfy the Cauchy distribution, For random variables that satisfy a Gaussian distribution, Standard deviation, and To dynamically adjust parameters.

[0028] Optionally, the learner of the Stacking ensemble model includes a lightweight gradient boosting machine algorithm, a limiting gradient boosting algorithm, a categorical feature boosting algorithm, and a random forest algorithm, wherein the lightweight gradient boosting machine algorithm, the limiting gradient boosting algorithm, the categorical feature boosting algorithm, and the random forest algorithm are used as base learners, and the limiting gradient boosting algorithm is used as a meta learner.

[0029] Secondly, the present invention provides a machine learning prediction system for rock shear strength parameters, characterized in that the machine learning prediction system for rock shear strength parameters includes: a data input device, a data output device, a processor, and a storage device, wherein the storage device includes a computer-readable storage medium storing a computer program, the computer program including program instructions, and the program instructions, when executed by the processor, cause the processor to implement the machine learning prediction method for rock shear strength parameters provided by the present invention.

[0030] In summary, the present invention has at least the following beneficial effects:

[0031] 1. This method selects P-wave velocity, rock density, uniaxial compressive strength and Brazilian tensile strength as predictive indicators of rock shear strength parameters, which has the advantages of low acquisition cost and short cycle;

[0032] 2. The Stacking ensemble model used in this method employs mature and high-performance algorithms in the field of machine learning, which can mine data features from multiple perspectives, give full play to their respective learning advantages, and improve prediction accuracy.

[0033] 3. This method uses multiple strategies to improve the sparrow search algorithm, and then optimizes the hyperparameters of the Stacking ensemble model. This can effectively avoid problems such as insufficient diversity during model population initialization, unstable search efficiency, and easy getting trapped in local optima.

[0034] 5. This method uses the SHAP method to analyze the interpretability of the prediction index of the rock shear strength prediction model, thereby improving the interpretability of the model;

[0035] 6. A system adapted to this method was constructed. This system is simple to operate, has excellent performance, and can improve the practicality of this method, and can be effectively applied to actual engineering projects. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating a machine learning method for predicting rock shear strength parameters according to an embodiment of the present invention.

[0038] Figure 2This is a ranking diagram of the importance of each input to cohesion in the rock shear strength prediction model of this invention.

[0039] Figure 3 This is a ranking diagram of the feature importance of each input to the internal friction angle in the rock shear strength prediction model of this invention embodiment;

[0040] Figure 4 This is a distribution diagram showing the characteristic influence of each input on cohesion in the rock shear strength prediction model of this invention.

[0041] Figure 5 This is a distribution diagram showing the characteristic influence of each input on the internal friction angle in the rock shear strength prediction model of this invention embodiment;

[0042] Figure 6 This is a characteristic dependence diagram of cohesion on P-wave velocity in an embodiment of the present invention;

[0043] Figure 7 This is a characteristic dependence diagram of the internal friction angle on rock density according to an embodiment of the present invention;

[0044] Figure 8 This is a schematic diagram of the framework of a machine learning prediction system for rock shear strength parameters according to an embodiment of the present invention;

[0045] Figure 9 This is a schematic diagram of the output interface of an embodiment of the present invention. Detailed Implementation

[0046] Specific embodiments of the present invention will now be described in detail. It should be noted that the embodiments described herein are for illustrative purposes only and are not intended to limit the invention. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be apparent to those skilled in the art that these specific details are not necessary to practice the invention. In other instances, well-known circuits, software, or methods have not been specifically described to avoid obscuring the invention.

[0047] Throughout this specification, references to "an embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in connection with that embodiment or example is included in at least one embodiment of the invention. Therefore, the phrases "in an embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily refer to the same embodiment or example. Furthermore, specific features, structures, or characteristics can be combined in one or more embodiments or examples in any suitable combination and / or sub-combination. Moreover, those skilled in the art will understand that the illustrations provided herein are for illustrative purposes and are not necessarily drawn to scale.

[0048] It should be noted in advance that, in one alternative embodiment, except for independent descriptions, the same symbols or letters appearing in all formulas have the same meaning and value.

[0049] In one optional embodiment, please refer to Figure 1 This invention provides a machine learning method for predicting rock shear strength parameters, the method comprising the following steps:

[0050] S1. Determine the prediction index and rock shear strength parameters, and establish a sample database.

[0051] Step S1 specifically includes the following steps:

[0052] S11. Determine the prediction index and shear strength parameters of rock.

[0053] Specifically, in this embodiment, the rock shear strength parameter prediction indicators include P-wave velocity, rock density, uniaxial compressive strength, and Brazilian tensile strength, while the rock shear strength parameters include cohesion and internal friction angle. Compared with other prediction indicators, the indicators selected in this embodiment have the advantages of low acquisition cost and short cycle time.

[0054] S12. Collect data on the predicted rock shear strength parameters and the rock shear strength parameters.

[0055] Specifically, in this embodiment, 199 sets of rock shear strength parameter prediction index data and rock shear strength parameter data were selected, including 49 sets of limestone, 50 sets of quartzite, 50 sets of slate, and 50 sets of quartz mica schist. Some of the data are shown in Table 1.

[0056] Table 1. Prediction index data and rock shear strength parameter data for some rocks.

[0057]

[0058] In Table 1, For P-wave velocity, Where is the rock density, UCS uniaxial compressive strength is , BTS is the Brazilian tensile strength, and c is the cohesion. It is the internal friction angle.

[0059] S13. Establish a sample database using the collected rock shear strength parameter prediction index data and the rock shear strength parameter data.

[0060] S2. Improve the sparrow search algorithm using an improved Tent chaotic map, dynamic adaptive weights, Levy flight strategy, and Cauchy-Gaussian hybrid mutation mechanism.

[0061] Step S2 specifically includes the following steps:

[0062] S21. Initialize the sparrow population using the improved Tent chaotic map.

[0063] Specifically, in this embodiment, the improved Tent chaotic mapping satisfies the following relationship:

[0064]

[0065] in, For the (i+1)th state value, Let N be the i-th state value, and N be the number of particles. The result is a random number between [0,1]. The improved Tent chaotic map solves the problem of small periodic points and unstable periodic points in the iterative sequence of the Tent chaotic map. Therefore, the initial population generated by the improved Tent chaotic map improves the distribution quality of the search space and strengthens the global search capability of the sparrow search algorithm.

[0066] S22. Introduce dynamic adaptive weights into the sparrow search algorithm to improve the discoverer position update formula.

[0067] Specifically, in this embodiment, a dynamic adaptive weight is introduced into the discoverer position update formula during the iterative optimization process in the sparrow search algorithm. The improved discoverer position update formula of the sparrow search algorithm is shown below:

[0068]

[0069] in, Let this be the position of the i-th sparrow in the j-th dimension during the (t+1)-th iteration. Let be the position of the i-th sparrow in the j-th dimension during the t-th iteration. For dynamic adaptive weights, Let be the global optimal solution for the sparrow population in the j-th dimension at the t-th iteration, and let Q be a random number following a normal distribution. This is a warning value. This is a safe value.

[0070] The dynamic adaptive weights specifically satisfy the following relationship:

[0071]

[0072] Where t is the number of iterations. ω represents the maximum number of iterations. Initially, ω has a large value, but it gradually decreases as the number of iterations increases, thus improving the convergence speed. Dynamic adaptive weights can alleviate the problem of the algorithm over-relying on individual discoverers, allowing the algorithm to better balance the relationship between global and local search, avoiding getting trapped in local optima.

[0073] S23. Follower position update formula based on the improved sparrow search algorithm of Levy flight strategy.

[0074] Specifically, in this embodiment, the Levy flight strategy can significantly enhance the global search capability of the sparrow search algorithm. When the sparrow search algorithm gets stuck in a local optimum, it can escape the local optimum by using a long-distance step size, thus solving the problem of the sparrow search algorithm easily getting stuck in local optima. The sparrow position update formula based on the Levy flight strategy is shown in the following equation:

[0075]

[0076] in, The updated sparrow position. The position of the sparrow before the update. These are step size control parameters. The step size is random.

[0077] The following relationship must be satisfied:

[0078]

[0079]

[0080] in, and The numbers are random numbers between 0 and 1, and all follow a normal distribution. is a constant and , for standard deviation This is a gamma function.

[0081] The improved sparrow search algorithm's follower position update formula is shown below:

[0082]

[0083] in, Let be the worst-case position of the sparrow population at the t-th iteration. For population size, This represents the optimal position of the sparrow population at the (t+1)th iteration.

[0084] S24. During the iterative operation of the sparrow search algorithm, the Cauchy-Gaussian hybrid mutation mechanism is used to perform adaptive mutation operations on the top S individuals with the best fitness values ​​in the current iteration step population, and the optimal individual is selected from the individuals before and after mutation to enter the next iteration step.

[0085] Specifically, in this embodiment, the Cauchy-Gaussian mixture mutation mechanism satisfies the following relationship:

[0086]

[0087] in, This is the position after the mutation. For the first The position of the optimal individual in the j-th dimension at the t-th iteration. For random variables that satisfy the Cauchy distribution, For random variables that satisfy a Gaussian distribution, Standard deviation, and To dynamically adjust parameters.

[0088] Standard deviation The following relationship must be satisfied:

[0089]

[0090] in, Let be the fitness value of the k-th optimal individual. For the first The fitness value of the optimal individual. It is a certain value, and .

[0091] In the Cauchygaussian mixture mutation mechanism In the early stages of iteration, larger values ​​are taken, allowing for the exploration of possible optimal solutions within a wider range of time intervals. In the later stages of iteration... It keeps decreasing, while The Gaussian mutation is continuously increased, which makes it easier for the algorithm to explore the optimal solution domain with a smaller variable time, thereby improving the algorithm's local exploration capability and improving convergence accuracy.

[0092] S3. Train the Stacking ensemble model using the sample database, and simultaneously optimize the hyperparameters in the Stacking ensemble model using an improved sparrow search algorithm to obtain a rock shear strength prediction model.

[0093] Specifically, in this embodiment, the sample database is divided into a training set, a validation set, and a test set in a ratio of 8:1:1 to complete the training, validation, and testing of the Stacking ensemble model. At the same time, an improved sparrow search algorithm is used to optimize the hyperparameters in the Stacking ensemble model, and finally a rock shear strength prediction model is obtained, which is represented by CISSA-Stacking.

[0094] More specifically, the learners in the Stacking ensemble model include the Lightweight Gradient Boosting Machine (LTGM), the Extreme Gradient Boosting (ATG), the Categorical Feature Boosting (CFS), and the Random Forest (RFS). The LTGM, AATG, CFS, and RFS serve as base learners, with the AATG as the meta-learner. Therefore, the hyperparameters of the Stacking ensemble model include: for the LTGM, the number of leaf nodes, learning rate, minimum leaf data size, feature sampling ratio, and L1 regularization coefficient; for the AATG, the maximum tree depth, learning rate, and number of weak learners; for the CFS, the tree depth, learning rate, and number of iterations; and for the RFS, the number of weak learners and the maximum tree depth.

[0095] The hyperparameter optimization results of the Stacking ensemble model in this embodiment are shown in Table 2:

[0096] Table 2. Hyperparameter optimization results of the Stacking ensemble model

[0097]

[0098] In Table 2, LightGBM represents the Lightweight Gradient Boosting Machine algorithm, XGBoost represents the Extreme Gradient Boosting algorithm, CatBoost represents the Categorical Feature Boosting algorithm, RF represents the Random Forest algorithm, and CISSA represents the Improved Sparrow Search algorithm.

[0099] Furthermore, to verify the performance of the rock shear strength prediction model, the coefficient of determination was used based on the sample database established in step S13. Root mean square error (RMSE) and mean absolute error (MAE) were used as evaluation indicators. Predictive models were constructed using LightGBM, XGBoost, CatBoost, RF, and the Stacking ensemble model (without CISSA optimization) and compared with the rock shear strength prediction model. The results are shown in Table 3.

[0100] Table 3 Comparison of prediction results from different models

[0101]

[0102] As shown in Table 3, in predicting c and At that time, CISSA-Stacking's R 2 The values ​​were 0.9936 and 0.9744, respectively; the RMSE values ​​were 0.4303 and 0.7635, respectively; and the MAE values ​​were 0.2161 and 0.5867, respectively. Overall, they significantly outperformed other models, demonstrating the superiority of CISSA-Stacking in predicting rock shear strength parameters.

[0103] Furthermore, to verify the effectiveness and practicality of CISSA-Stacking, it was applied to predict the shear strength parameters of five types of disturbed rocks in a mining area. The measured data for these five types of disturbed rocks in the mining area are shown in Table 4.

[0104] Table 4. Measured results of various data of rocks in a certain mining area

[0105]

[0106] The rock shear strength parameters of this mining area were predicted using CISSA-Stacking, and the absolute error (AE) evaluation model was used to predict the error. The prediction results are shown in Table 5.

[0107] Table 4. Prediction results of rock shear strength parameters in a certain mining area

[0108]

[0109] Table 5 shows that CISSA-Stacking is effective in predicting the c and t values ​​of five types of disturbed rocks. When the AE was below 0.6, it proved that the results predicted by CISSA-Stacking were close to the true values ​​and could be effectively applied to practical engineering.

[0110] S4. Use the rock shear strength prediction model to predict the rock shear strength parameters, and use the SHAP interpretability method to evaluate the contribution of each rock shear strength parameter prediction index to the prediction results.

[0111] Specifically, in this embodiment, feature importance analysis is performed on the four inputs of the trained rock shear strength prediction model. The absolute value of the average Shapley value for each input in the rock shear strength prediction model is calculated using SHAP to determine the importance of each rock shear strength parameter prediction index. The feature importance ranking diagrams for each input of the rock shear strength prediction model to cohesion and internal friction angle in this embodiment are shown below. Figure 2 and Figure 3 As shown. By Figure 2It can be seen that, for cohesion, the input with the highest absolute value of the Shapley value and the most significant feature importance is... The next two are UCS, BTS, and .Depend on Figure 3 It can be seen that, for the internal friction angle, the input with the highest absolute value of the Shapley value and the most significant feature importance is... The next two are UCS, And BTS.

[0112] Furthermore, a feature influence distribution analysis was performed on the four inputs of the trained rock shear strength prediction model, resulting in the feature influence distribution diagrams of each input on cohesion and internal friction angle, as shown below. Figure 4 and Figure 5 As shown. By Figure 4 It can be seen that, regarding cohesion, The impact on the prediction results is most significant when A larger value will have a positive contribution to the prediction results. Figure 5 It can be seen that, regarding the internal friction angle, The impact on the prediction results is most significant when Larger values, or those close to the median, will positively contribute to the prediction results. Furthermore, from... Figure 4 and Figure 5 It can also be seen that, for the four rock shear strength prediction indices, as they increase, the Shapley values ​​of cohesion and internal friction angle also show a gradual increasing trend, proving that the four rock shear strength prediction indices all show a relatively significant positive correlation with the rock shear strength prediction results.

[0113] Furthermore, to evaluate the interaction among the rock shear strength parameter prediction indices during the prediction process, the SHAP method was used to quantify the dependencies among different rock shear strength parameter prediction indices, thereby analyzing the influence mechanism of the four rock shear strength parameter prediction indices on the prediction results. In this embodiment, the rock shear strength parameter prediction index with the most significant impact was selected, and a feature dependency analysis was performed with another rock shear strength parameter prediction index that interacted most significantly with it. The resulting partial feature dependency diagrams are shown below. Figure 6 and Figure 7 As shown. By Figure 6 It can be seen that the factor that contributes the most to the cohesion prediction results is... In other words, with With the increase of [unclear], both the Shapley value and UCS showed an upward trend, proving that UCS has an effect on [unclear]. The impact is quite significant. Figure 7It can be seen that the factor that contributes the most to the prediction of the internal friction angle is... In other words, with With the increase of Shapley value, UCS does not show an upward trend, proving that UCS is important for... The impact was not significant.

[0114] It should be noted that in some cases, the actions described in the specification can be performed in different orders and still achieve the desired results. In this embodiment, the order of steps is given only to make the embodiment clearer and easier to explain, and not to limit it.

[0115] In one optional embodiment, please refer to Figure 8 To improve the practicality of this method, the present invention also provides a machine learning prediction system for rock shear strength parameters. The machine learning prediction system for rock shear strength parameters includes: a data input device 1, a data output device 2, a processor 3, and a storage device 4. The storage device 4 includes a computer-readable storage medium storing a computer program. The computer program includes program instructions, which, when executed by the processor 3, cause the processor 3 to perform the contents described in steps S1 to S4.

[0116] Specifically, in this embodiment, data input device 1 may include a mouse and keyboard, data output device 2 may be a computer monitor, processor 3 may be a computer processor, and storage 4 may be a computer storage device. Personnel can use data input device 1 to input a total of 199 sets of rock shear strength parameter prediction index data and rock shear strength parameter data into the system to construct a sample database, which is then used for model training to obtain a rock shear strength prediction model. Both the sample database and the rock shear strength prediction model are stored in storage 4.

[0117] More specifically, when relevant personnel input a set of predicted rock shear strength parameters into the system, processor 3 uses the rock shear strength prediction model to predict cohesion and internal friction angle, and displays the prediction results and the input predicted rock shear strength parameters on data output device 2. For example... Figure 9 As shown, when the input P-wave velocity, rock density, uniaxial compressive strength, and Brazilian tensile strength are 3716.7 m·s⁻¹, the rock density, uniaxial compressive strength, and Brazilian tensile strength are respectively... -1 2.718 g·cm -3 At 113.1 MPa and 6.3 MPa, the predicted cohesion and internal friction angle are 17.55 MPa and 41.27°, respectively.

[0118] In summary, the present invention has at least the following beneficial effects:

[0119] 1. This method selects P-wave velocity, rock density, uniaxial compressive strength and Brazilian tensile strength as predictive indicators of rock shear strength parameters, which has the advantages of low acquisition cost and short cycle;

[0120] 2. The Stacking ensemble model used in this method employs mature and high-performance algorithms in the field of machine learning, which can mine data features from multiple perspectives, give full play to their respective learning advantages, and improve prediction accuracy.

[0121] 3. This method uses multiple strategies to improve the sparrow search algorithm, and then optimizes the hyperparameters of the Stacking ensemble model. This can effectively avoid problems such as insufficient diversity during model population initialization, unstable search efficiency, and easy getting trapped in local optima.

[0122] 5. This method uses the SHAP method to analyze the interpretability of the prediction index of the rock shear strength prediction model, thereby improving the interpretability of the model;

[0123] 6. A system adapted to this method was constructed. This system is simple to operate, has excellent performance, and can improve the practicality of this method, and can be effectively applied to actual engineering projects.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A machine learning method for predicting rock shear strength parameters, characterized in that, Includes the following steps: Determine the prediction index and parameters of rock shear strength, and establish a sample database; Initialize the sparrow population using an improved Tent chaotic map; The improved Tent chaotic mapping satisfies the following relationship: in, For the (i+1)th state value, Let N be the i-th state value, and N be the number of particles. A random number between [0, 1]; Introduce dynamic adaptive weights into the sparrow search algorithm to improve the finder position update formula; The improved sparrow search algorithm's discoverer position update formula is shown below: in, Let this be the position of the i-th sparrow in the j-th dimension during the (t+1)-th iteration. Let be the position of the i-th sparrow in the j-th dimension during the t-th iteration. For the dynamic adaptive weights, Let be the globally optimal solution for the sparrow population in the j-th dimension at the t-th iteration. Q For random numbers that follow a normal distribution, This is a warning value. This is a safe value; Follower position update formula based on Levy flight strategy to improve sparrow search algorithm; The improved sparrow search algorithm's follower position update formula is shown below: in, Let be the worst-case position of the sparrow population at the t-th iteration. For population size, This represents the optimal position of the sparrow population in the (t+1)th iteration. The step size is random. During the iterative operation of the sparrow search algorithm, the Cauchy-Gaussian hybrid mutation mechanism is used to perform adaptive mutation operations on the top S best individuals with the best fitness values ​​in the current iteration step population, and the best individual is selected from the individuals before and after mutation to enter the next iteration step. The Cauchy-Gaussian mixture mutation mechanism satisfies the following relationship: in, This is the position after the mutation. For the first The position of the optimal individual in the j-th dimension at the t-th iteration. For random variables that satisfy the Cauchy distribution, For random variables that satisfy a Gaussian distribution, Standard deviation, and To dynamically adjust parameters; The Stacking ensemble model was trained using the sample database, and the hyperparameters in the Stacking ensemble model were optimized using an improved sparrow search algorithm to obtain a rock shear strength prediction model. The rock shear strength prediction model is used to predict the rock shear strength parameters, and the SHAP interpretability method is used to evaluate the contribution of each rock shear strength parameter prediction index to the prediction results.

2. The machine learning prediction method for rock shear strength parameters according to claim 1, characterized in that, The process of determining the predictive index of rock shear strength parameters and the rock shear strength parameters, and establishing a sample database, includes the following steps: Determine the prediction index and shear strength parameters of rock; Collect data on predicted rock shear strength parameters and rock shear strength parameters; A sample database was established using the collected rock shear strength parameter prediction index data and the rock shear strength parameter data.

3. The machine learning prediction method for rock shear strength parameters according to claim 1, characterized in that: The predicted rock shear strength parameters include P-wave velocity, rock density, uniaxial compressive strength, and Brazilian tensile strength. The rock shear strength parameters include cohesion and internal friction angle.

4. The machine learning prediction method for rock shear strength parameters according to claim 1, characterized in that: The learners of the Stacking ensemble model include the Lightweight Gradient Boosting Machine (LTGM), the Extreme Gradient Boosting (TGP), the Categorical Feature Boosting (CFP), and the Random Forest (RFR). The LTGM, TGP, CFP, and RFR serve as base learners, and the TGP serves as the meta learner.

5. A machine learning prediction system for rock shear strength parameters, characterized in that, The rock shear strength parameter machine learning prediction system includes: a data input device, a data output device, a processor, and a storage device. The storage device includes a computer-readable storage medium storing a computer program. The computer program includes program instructions, which, when executed by the processor, cause the processor to implement the rock shear strength parameter machine learning prediction method as described in any one of claims 1-4.

Images